Studying an element of the information search system: The choice process approximated to the natural intelligence

Abstract

Nowadays the task of studying the information search as a system is relevant, one of its elements being the process of choice. The study is essential for automating and performing subsequent statistical analysis of the process, so the logical, structural and functional models of the process have been obtained and studied. In the study, the authors used the deductive system (developed by them) of the process of choosing the best alternative. First of all, the network modelling method was applied, alongside with the graphic way of calculation and formulae for defining the early and late event occurrence term. It has been found that the optimal time implies one cognitive work on selecting the best alternative from many known ones proceeding from a certain volume of knowledge, which needs automating. Next, Kolmogorov equations for the choice process were generated, with data obtained and a graph built which allowed making some useful conclusions. Having only a half of the amount of knowledge in a subject area enables one to get a simple intelligent system repeating the critical path 1-2-3-4-5. A smaller volume of knowledge can be used for a learning intelligent system, and a greater volume – in an intelligent system featuring error correction and check for compliance with the expected result. Then, an intelligent system conducting exploratory research has to possess the total volume of knowledge in order to maximize its cognitive power up to k = 1, in which case the efficiency of this system will be comparable to that of the natural intelligence. This is why the authors looked for an approach which would be close to the natural intelligence in representing metaknowledge. The paper provides a specific representation of metaknowledge, with the decision tree being that of a multitude of precedents. All the rules are applied for obtaining the tree, in which a consequent is a corresponding antecedent, and the transition to it corresponds to the obtained tree structure. Such trees can be used for representing metaknowledge and searching for solutions while organizing the efficient choice process.